돼지의 3차원 메쉬로부터 다층 퍼셉트론 기반 무게 예측 방법

Abstract

Maintaining the quality of livestock is important to ensure competitiveness in the livestock industry. For this purpose, smart farm construction is being actively carried out. Weight and body type are important factors for growth management of livestock. In this study, we pro- pose a method of predicting the weight of a pig from shape information obtained from a 3D measurement system. The shape consists of triangular elements, and the pig is divided into five regions to calculate the main dimensions. After calculation of the widest position of the pig, the positions of the fore legs, hind legs, and torso are defined. A model for weight estima- tion is created by using MLP (multi-layer perceptron) to learn the calculated dimensions and the actual weight. As a result of applying the learning model, accurate prediction values were obtained.